Human observer functions x(λ) y(λ) z(λ) are response functions used to characterize colors perceived by the human eye. The existing Commission Internationale de l'Eclairage (CIE) and International Organization for Standardization (ISO) definitions of the human observer functions are the basis for color management systems and for standard profile formats such as those defined by the International Color Consortium (ICC). The fact that the ICC has been embraced and incorporated into operating systems such as Apple™ OS X and applications such as Adobe™ PhotoShop™ is evidence that the standard is reasonably useful and effective for ensuring or at least improving the quality of color appearance of images. The paradigm of conversions from “source” to “destination” implies preserving quality and/or appearance of digital color images.
Standards committees such as IDEAlliance have defined very clear measured color target values and tolerances using (CIEXYZ) and (CIELAB) in order to ensure that the visual appearance of hard copy proofing systems from multiple venders are consistent for a given digital input image file. Assessments of prints generated by multiple venders using these standards have been performed on numerous occasions with great success both within committee meetings and at public events to showcase the efficacy of these standards. In most cases, venders make use of the ICC standard or similar colorimetrically-based formats in order to ensure correct output of printed images.
Although existing CIE and ISO standards appear to work well for multiple prints viewed under the same illumination, achieving the same level of standardization for soft proofing has proven to be more challenging. An example of this reality is the fact that the calculations for CIELAB in the IDEAlliance certification of soft proofing systems are normalized to the white point of the display rather than to D50. The reason for this non-standard calculation of CIELAB is the widespread practice of adjusting the white point of a display in order to match visually the white balance of the D50 illumination used to view the corresponding print. In theory, it should be adequate to measure the white balance of the illumination and merely ensure that white point of the display is adjusted to match the illumination according to measured values.
In reality, an offset to the measured white is often applied in order to ensure a good visual match between display and print. However, it has been observed that it is difficult to ensure both accurate appearance of white/gray balance and accurate appearance of other critical regions of color such as skin tones. If there exists an error in the standard, one would expect a simple correction would be sufficient.
The need for a slight modification or offset to the white point appears to have some validity as indicated by scientific studies that have been performed in the area of color matching functions determined via the “Maxwell method,” i.e. the matching of whites, vs. the more common “saturation method,” which determines color matching functions based on matching saturated colors. For example, in section 5.6.6 of Color Science: Concepts and Methods, Wyscecki summarizes the work done in the area of comparing the saturation and Maxwell methods. At the end of the section, Wyscecki states, “The deviations are appropriately termed failure of the additivity law of color matching in a bipartite field. Suggestions have been made with regard to possible causes of these failures. They include chromatic adaptation (Crawford, 1965), Maxwell spot (Palmer, 1980), and interactions or linkages between different cone mechanisms (Ingling and Drum, 1973), but further work is obviously needed to resolve the conundrum.”
Likewise, in a series of papers I-VI entitled “Toward a more accurate and extensible colorimetry,” Thornton describes various experiments involving the matching of whites. Converting his data to units of CIELAB indicate a disagreement between observed matches of different white spectra and predicted matches of up to 25 ΔE as calculated using the standard human observer. In a similar study published in 1993 by North and Fairchild, “Measuring Color-Matching Functions. Part II” (1993, North and Fairchild) Thornton's observations appeared to be confirmed, although the analysis of the results was performed based on the model of the deviate human observer, i.e. was explained by the differences between individual observers rather than interpreted to imply a correction to existing standard human observers functions.
Most recently at the Society for Imaging Science and Technology (IS&T) Color Imaging Conference 16 in Portland, Oreg., a presentation, “Color Vision and More Comprehensive Color Appearance Models” was given by Hirohisa Yaguchi, Chiba University (Japan). In this presentation, Yaguchi-san showed plots of color matching functions obtained using his own eyes on his color matching apparatus using first the saturation and then the Maxwell method. He pointed out that the plots contained differences that were not insignificant, although no further explanation was given.
Virtual proofing systems are required to display an accurate match between display and hard copy with little effort from the user. If one accounts for metamerism and for errors in conventional XYZ calculations, one finds that the overall hues and white balance are reasonable. For critical color requirements, however, there can exist unacceptable visual differences between display and hard copy in skin tones (for example) that equate to several percentage points of change in magenta dot gain. Although this error is small, it is enough to be problematic for critical work.
Therefore there remains a need for an improvement to conventional XYZ calculations. It is noted that the “deviant observer” described by Fairchild does vary or modify the x(λ) y(λ) z(λ) functions with size of the color observed (characterized as an angle indicating the size of the cone of viewing from the eye to a circle indicating the region of the color being observed), yellowing of the lens of the eye due to age, effect of the macula of the eye, etc. These adjustments endeavor to account for differences due to size of color and to account for observer to observer differences. However, the particular observer functions for a given size of color and particular individual is the same regardless of the spectral power distribution (SPD) S(λ) of the color stimulus being observed. Reference is made to such observer functions as “static” meaning they do not change with SPD S(λ). By contrast, reference is made to human observer functions that change or adapt depending on the SPD S(λ) of the particular color being viewed as “non-static” meaning that they do vary with S(λ).
If the human observer functions were static with S(λ), pairs of colors would always match as long as the calculated values of XYZ were nearly the same, regardless of whether the colors were very saturated or very neutral. However, if the human observer functions are not static with S(λ), it is quite possible for saturated color pairs to match but for neutral pairs of colors to appear different even though their XYZ's were the same. By allowing the observer functions to be non-static with S(λ), discrepancies between saturated and neutral regions of color can be resolved, thereby ensuring that pairs of color will always match when their XYZ's are similar regardless of their S(λ) characteristics.
The observer functions are the basis for color management systems. A review how color management systems work in relation to the XYZ values calculated using the observer functions, and L*a*b* values derived in turn from XYZ is necessary, with the understanding that any improvements or changes to the observer functions will correspondingly affect XYZ, L*a*b*, and the functionality and efficacy of color management systems that rely on these color values.
Current color management systems such as Apple ColorSync™ convert color pixel data from source device dependent coordinates to destination device dependent coordinates via a device independent color space referred to as the Profile Connecting Space or PCS. The PCS is generally based upon tristimulus values CIEXYZ or perceptually more uniform color spaces such as CIELAB which are calculated from CIEXYZ.
The general procedure is to define source colors either as a list of PCS values (such as the Pantone™ library) or as a set of device pixel values such as CMYK with an associated profile or characterization. The most common format for such profiles is that of the International Color Consortium (ICC). In the latter case, the processing engine of the system known as the color matching module or method (CMM) converts the device dependent values (e.g. CMYK) to an appropriate PCS (e.g. CIELAB or CIEXYZ) via interpolation of tables within the ICC profile for the device.
In order for these PCS values to demonstrate usefulness, the color management system is usually invoked for the purpose of converting these PCS values (once they are fetched or calculated) to a device dependent destination such as an RGB monitor or digital projector. This is performed via the color matching method (CMM) in conjunction with the ICC profile of the destination device. Implicit in the above system is the assumption that at some point prior to the above conversion, color measurements have been performed in order to calculate the PCS values. For individual colors in a library such as Pantone™, this is the only effort that is required. In the case of a complex device requiring an ICC profile, multiple color measurements are performed and high resolution look up tables are calculated from that data in order to construct a reasonable approximation to the color characteristics of the device, for example, to estimate reasonable expected measured values of L*a*b* or XYZ for any combination of CMYK pixel values. Likewise, for complex devices, an inverse table is constructed and stored in the ICC profile to convert PCS values to the devices codes of the device, e.g. RGB or CMYK.
All the above requires a means for measuring and quantifying color. The two most common methods are spectral measurement and calorimetric measurement. Spectral measurement results in spectral data, either in the form of a reflectance spectrum R(λ) or the spectral power distribution (SPD) S(λ). The former is used for reflective materials while the latter is used for an emissive device such as a monitor. Since the SPD is required to calculate XYZ using the observer functions, the SPD for reflective materials is usually calculated by multiplying R(λ) by the SPD of the illuminant I(λ).
The present invention modifies the behavior of the human observer functions as they vary with λ based on changes of location in color space of the color being measured. In particular, colors in the region of white and gray will require somewhat different observer functions from colors that are saturated. These modifications to the calculation of XYZ from S(λ) will affect the characterizations and the conversions of colors processed by a color management system. In particular, the physical outcome of this invention will be that images reproduced on a display or digital projector using a color management system will measure differently from the current systems particularly in regions of neutral and white.
Furthermore, for systems comprising extremely narrow band primaries, such as a projector using RGB lasers, there will be a significant improvement in the color reproduction of original color images. For example if the original image is displayed on a monitor, converted with the color management system, and displayed using a projector with RGB lasers, the visual match between the two images will be significantly improved particularly in regions of white and gray if the color management system is modified to use the improved human observer functions.